Microsoft releases the Microsoft Agent Framework The (public preview) is an open-source SDK that combines AutoGen’s (agent runtime, multi-agent pattern) and Semantic Kernel’s (enterprise control, state, plug-ins) core ideas to enable teams to build, deploy and monitor production-grade AI agent and multiagent workflows. The framework can be used with Python or.NET. It integrates seamlessly into Azure AI Foundry’s Agent Service Scaling and Operations
Microsoft Ships What?
- The API and agent runtime surface is consolidated. The Agent Framework is a continuation of AutoGen, which provides single- and multiple-agent abstractions. It also adds Semantic Kernel enterprise features such as thread-based state-management, type safety and filters. Microsoft presents it as a successor, built by the teams that worked on both projects.
- The first-class orchestration modes. This is a great way to support Agent Orchestration (LLM-driven decision-making) and Workflow orchestration (deterministic and business-logic flows multi-agent), allows hybrid systems to coexist creative planning with handoffs that are reliable, as well as constraints.
- Interoperability of platforms and pro-code. The basis
AIAgentThe interface was designed to allow users to switch chat models and interoperate between Azure AI Foundry Agents and OpenAI Assistants. This reduces vendor lock-in on the application level. - Open-source, multi-language SDKs under MIT license. It is important to note that the word “you” means “you”. GitHub repo AutoGen continues to be maintained (bug fixes, security patches) with guidance on using Agent Framework for new builds. AutoGen is still maintained with bug fixes and security patches, but the recommendation to use Agent Framework instead for new builds.
When is it produced?
Azure AI Foundry’s Agent Service It provides the managed runtime. This includes linking models, frameworks, and tools; managing thread state, enforcing content safety, identity, and wired in observability. This also includes Orchestration by multiple agents It targets complex enterprise scenarios and is different from Copilot Studio’s low-code solution.
But how is it connected to ‘AI economics’?
Enterprise AI is primarily driven by the token throughput and latency. Failure recovery, observability, and failure recovery are also important. Microsoft’s consolidation addresses those by (a) giving one runtime abstraction for agent collaboration and tool use, (b) attaching production controls—telemetry, filters, identity/networking, safety—to the same abstraction, and (c) deploying onto a managed service that handles scaling, policy, and diagnostics. It reduces the amount of data that needs to be stored. “glue code” The model catalog + toolchain of Azure AI Foundry aligns to the typical cost and fragility in multi-agent system.
Architectural Notes and Developer Surface
- Runtime & state: Agents coordinate via a runtime that handles lifecycles, identities, communication, and security boundaries—concepts inherited and formalized from AutoGen. Threads The unit of State, which allows for reproducible runs, retries and audits.
- Functions & plugins: It relies heavily on the Semantic Kernel plugin architecture. (
- Model/provider flexibility: The same interface is available for Azure OpenAI (OpenAI), local runtimes like Ollama/Foundry Local and GitHub Models. Cost/performance can be adjusted per task by rewriting only the orchestration logic.
Enterprise context
Microsoft frames the release As part of an overall push towards interoperable and standard-friendly “agentic” systems across Azure AI Foundry—consistent with prior statements about multi-agent collaboration, memory, and structured retrieval. As Foundry governance and observability controls stabilize, expect tighter connections.
We like this direction because it collapses two divergent stacks—AutoGen’s multi-agent runtime and Semantic Kernel’s enterprise plumbing—into one API surface with a managed path to production. Azure AI Foundry Agent Service handles identity, tool orchestration, and content security, allowing teams to focus on policy instead of glue code. The Python/.NET equality and the flexibility provided by providers such as Azure OpenAI (OpenAI), GitHub Models or local runtimes make it possible to tune cost and performance without rewriting an orchestration.
Asif Razzaq, CEO of Marktechpost Media Inc. is a visionary engineer and entrepreneur who is dedicated to harnessing Artificial Intelligence’s potential for the social good. Marktechpost is his latest venture, a media platform that focuses on Artificial Intelligence. It is known for providing in-depth news coverage about machine learning, deep learning, and other topics. The content is technically accurate and easy to understand by an audience of all backgrounds. This platform has over 2,000,000 monthly views which shows its popularity.

